on visual and linguistic information

Brief summary of the paper:
There are two competing paradigms: one argues that the continuous linguistic input are encapsulated from other cognitive and perceptual systems, while the perceptual systems are also autonomous modules; the other argues the nonlinguistic information influence the syntactic processing, i.e. the perceptual and syntactic processes intervene. The experiment presented in this paper support the second argument. The experiment is designed with two settings varied by two variables: one-/two-referent context vs. ambiguous/unambiguous instruction, and people's eye movement is measured as the performance of comprehension of the given instructions. The null hypothesis is that if initial syntactic processing is encapsulated (the first paradigm), there should be no significant difference between the two results, while the experiment does show the different fixation patterns.

My queries:
It was a well-designed experiment--using non-linguistic information (eye movement) to examine the link between visual context and temporal ambiguities given by syntax structure. However, I found it is somehow dangerous to claim "visual context affected the resolution of temporary ambiguities". In fact, there are two types of ambiguities in the experiment: visual (which apple) and temporal (which destination). If considering the visual ambiguity, it seems that people perform the best when there's no ambiguity at all, they perform well when the given information help to resolve the ambiguity, and the worst when there is no (visual) ambiguity in the context but the given information introduce another (temporal) ambiguity. If there is a third setting, which remove the empty towel from the second setting, my conjecture is that the result will be similar to the second setting.